Technology is not neutral. It carries the values of its creators and the interests of those who fund it.

Last week I wrote that we cannot yet fully define the problem AI will solve, and that the ecosystem assembling around AI matters as much as the tool itself. Both ideas point to the same truth: we are building the plane while flying it.

As Klaus Schwab put it in 2016: "The Fourth Industrial Revolution will change not only what we do but also who we are."

On one level that means AI diagnosing cancer, gene editing, brain-computer interfaces. But there is another dimension inside every organisation: the fusion of human and machine intelligence in how we work, decide and collaborate. Teams that include both human and artificial agents. Cultures that have to be deliberately designed to protect what is distinctly human about the way we work.

But technology is not neutral and it carries the values of its creators and the interests of those who fund it. An AI recruitment tool trained on historical data does not just reflect past decisions, it scales them. A credit scoring algorithm does not just assess risk, it embeds existing inequality into every future decision. A content recommendation algorithm does not just surface what you want to see, it shapes what you believe is true.

Which is why ethics cannot wait for an audit, and why 2026 is the year to get this right. There are seven live challenges every AI deployment sits inside, often several at once: privacy and surveillance, algorithmic bias, inequality and the digital divide, unemployment and job displacement, sustainability and the energy cost of AI, the AI black box, and trust and misuse.

Most organisations treat these as edge cases, but they are not. They are the default conditions of operating with AI at scale.

From August 2026 (just three months away, pending final legislative agreement) the EU AI Act makes several of them legally enforceable for the first time. The fines are significant: up to €35 million or 7% of global turnover for prohibited practices, up to €15 million or 3% for high-risk non-compliance and up to €7.5 million or 1.5% of global turnover for providing false or misleading information.

A tangible example most technology leaders will recognise: if your organisation uses AI to screen job candidates, that system is now classified as high-risk under the Act. The obligations include documented risk management, human oversight, transparency about how decisions are made, and ongoing monitoring. Non-compliance carries real financial consequences.

The organisations that will navigate this well are not the ones scrambling to comply in August. They are the ones that asked the right questions before they deployed.

What question are you least comfortable answering about your current AI deployments?